ClusterRAG: Cluster-Based Collaborative Filtering for Personalized Retrieval-Augmented Generation
The paper introduces ClusterRAG, a novel approach for Personalized Retrieval-Augmented Generation (RAG). It emphasizes the importance of collaborative signals from similar users to improve document retrieval and generation. Extensive experiments demonstrate that ClusterRAG outperforms existing methods by leveraging user profiles and clustering techniques.
- ▪ClusterRAG utilizes density-based clustering to organize users into semantically coherent groups.
- ▪The method performs retrieval at both the cluster and document levels, enhancing personalization.
- ▪Experiments on the LaMP benchmark show that ClusterRAG consistently yields superior performance across various tasks.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Information Retrieval arXiv:2605.18769 (cs) [Submitted on 14 Apr 2026] Title:ClusterRAG: Cluster-Based Collaborative Filtering for Personalized Retrieval-Augmented Generation Authors:Gibson Nkhata, Uttamasha Anjally Oyshi, Quan Mai, Susan Gauch View a PDF of the paper titled ClusterRAG: Cluster-Based Collaborative Filtering for Personalized Retrieval-Augmented Generation, by Gibson Nkhata and 3 other authors View PDF HTML (experimental) Abstract:Personalized Retrieval-Augmented Generation (RAG) relies on accurately selecting user-relevant documents. In practice, existing RAG approaches often suffer from high retrieval costs and overlook that collaborative signals from similar users can enhance personalized generation for the current user.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
